0% found this document useful (0 votes)
95 views13 pagesSelinker
error correction analysis
Uploaded by
Kruno MedakCopyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF or read online on Scribd
0% found this document useful (0 votes)
95 views13 pagesSelinker
error correction analysis
Uploaded by
Kruno MedakCopyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF or read online on Scribd
/ 13
30 PART Two.
approximative system for interlanguage. This term has the advantage
of implying the developmental nature of language learning, since
the learners’ system is continually being modified as new clements
are incorporated throughout the learning process. Such developing
systems are evident in learner’s errors. Nemser notes that the atten-
tion to the content rather than the form of utterances by the learner
(cf. communication strategy) seen in learners’ simplified grammar,
may also be characteristic of the teacher's speech (cf. Ferguson,
1971). He gives phonological examples of the intermediate and often
independent systems developed by the learner. Such systems have
internal coherence and are not simply corrupt versions of the target
language or mother tongue and their characteristics change accord-
ing to the degree of learning. Such systems are illegitimates socio-
linguistically —their users do not normally form speech communities
(cf. Richards, this section)—though they are not without socio-
linguistic significance, as Nemser illustrates.
In the third paper I have tried to suggest how our view of the
learner's developing system is determined by the social relationships
holding between the learner and the target language community.
‘An attempt is made to relate the linguistic concepts developed else-
here in this volume (interlanguage, approximative system, over-
generalization, communication strategy, etc.) to an overall view of
the different social settings in which the learning of English takes
place. The non-linguistic dimensions of language learning illustrated
include the effects of the perception of the target language community
by the Jeamer, the structure and rigidity of social roles in the
community, the economic and social strength of the learners if they
form a community, and the roles for which English is used by the
learner. An analysis of the learner’s developing linguistic system can
‘be complemented by consideration of the social factors which give
these features significance. The paper hence makes distinctions
between second language and foreign language contexts for the
learning of English which have important consequences for our
view of the learner's interlanguage or developing system,
3
Interlanguage
LARRY SELINKER
Reprinted from IRAL, Vol. X/3, 1972, published by Julius Groos
Verlag, Heidelberg.
1 Introduction
This paper! discusses some theoretical preliminaries for researchers
concerned with the linguistic aspects of the psychology of second
language learning. These theoretical preliminaries are important
because without them itis virtually impossible to decide what data
are relevant to a psycholinguistic theory of second language
learning.
Ii also important to distinguish between a teaching perspective
and a learning one. As regards the ‘teaching’ perspective, one might
very well write a methodology paper which would relate desired
output to known inputs in a principled way, prescribing what has
to be done by the teacher in order to help the learner achieve learn-
ing. As regards the ‘learning’ perspective, one might very well write
a paper describing the process of attempted learning of a second
language, successful or not; teaching, textbooks, and other ‘external
‘aids’ would constitute one, but only one, important set of relevant
variables. In distinguishing between the two perspectives,? claims
about the internal structures and processes of the learning organism
take on a very secondary character in the teaching perspective; such
claims may not even be desirable here. But such claims do provide
the raison d'étre for viewing second language learning from the
learning perspective. This paper is written from the learning per-
spective, regardless of one’s failure or success in the attempted
learning of a second language.
In the learning perspective, what would constitute the psycho-
logically relevant data of second language learning? My own
Position is that such data would be those behavioral events which
would lead to an understanding of the psycholinguistic structures
and processes underlying ‘attempted meaningful performance’ in a
31 32. L. SELINKER
second language. The term ‘meaningful performance’ will be used
here to refer to the situation where an ‘adult™ attempts to express
‘meanings, which he may already have, in a language which he is in
the process of learning. Since performance of drills in a second
language classroom is, by definition, not meaningful performance,
it follows that from a learning perspective, such performance is,
in the long run, of minor interest. Also, behavior which occurs in
experiments, using nonsense syllables fits into the same category and
for the same reason. Thus, data resulting from these latter behavioral
situations are of doubtful relevancy to meaningful performance
situations, and thus to a theory of second language learning.
It has Jong seemed to me that one of our greatest difficulties in
establishing a psychology of second language learning which is
relevant to the way people actually learn second languages, has been
our inability ¢o identify unambiguously the phenomena we wish to
study. Out of the great conglomeration of second language be-
havioral events, what criteria and constructs should be used to
establish the class of those events which are to count as relevant in
theory construction’ One set of these behavioral events which has
elicited considerable interest is the regular reappearance in second
language performance of linguistic phenomena which were thought
to be eradicated in the performance of the learner. A correct under-
standing of this phenomenon leads to the postulation of certain
theoretical constructs, many of which have been set up to deal with
other problems in the field. But they also help clarify the phenomenon
under discussion, These constructs, in turn, give us a framework
within which we can begin to isolate the psychologically relevant data
of second language learning. The new perspective which an examina-
tion of this phenomenon gives us is thus very helpful both in an
identification of relevant data and in the formulation of a psycho-
linguistic theory of second language learning. The main motivation
for this paper is the belief that itis particularly inthis area that pro-
‘gress can be made at this time.
2 Interlanguage’ and latent structures
Relevant behavioral events in a psychology of second fanguage
learning should be made identifiable with the aid of theoretical
constructs which assume the major features of the psychological
structure of an adult whenever he attempts to understand second
TCpeds =i LAD oes
INTERLANGUAGE 33,
fanguage sentences or to produce them. If, in a psychology of
second language learning, our goal is explanation of some important
aspects of this psychological structure, then it seems to me that we
are concered in large part with how bilinguals make what Wein-
reich (1953, p, 7) has called ‘interlingual identifications’. In his book
Languages in Contact, Weinreich discusses—though briefly —the
practical need for assuming in studies of bilingualism that such
identifications as that of a phoneme in two languages, or that of a
grammatical relationship in two languages, or that of a semantic
feature in two languages, have been made by the individual in ques-
sion in a language contact situation. Although Weinreich takes up
many linguistic and some psychological questions, be leaves com-
pletely open questions regarding the psychological structures within
‘hich he assumes ‘interlingual identifications’ exist; we assume that
there is such a psychological structure and that itis latent in the brain,
activated when one attempls to learn a second language,
closest thing in the literature to the concept latent psycho-
Togical structure is the concept of latent language structure (Lenne-
berg, 1967, especially pp. 374-379) which, according to Lenneberg,
(a) is an already formulated arrangement in the brain, (b) is the
biological counterpart to universal grammar, and (c)is transformed
\ by the infant into the realized structure of a particular grammar in
accordance with certain maturational stages JFor Bie purposes of”
this paper, I will assume the existence of the latent structure described
‘by Lenneberg, I shall further assume that there exists in the brain an
already formulated arrangement which for most people is different
from and exists in addition to Lenneberg's latent language structure.
It is important to state that with the latent structure described in this
Paper as compared to Lenneberg's, there is no genetic timetable;
there is no direct counterpart to any grammatical concept such as
‘universal grammar’; there is no guarantee that the latent structure
Will be activated at all, there is no guarantee that the latent structure
Will be ‘realized’ into the actual structure of any natural language
(ie. there is no guarantee that attempted learning will prove
successful), and there is every possibility that an overlapping exists
between this latent language acquisition structure and other
intellectual structures. es
~The crucial assumption we are making here is that those adults
} Who ‘succeed’ in learning a second language so that they achieve
( native-speaker ‘competence’ have somehow reactivated. the latent
Nanguage srt whi i absolute
coppatin » Wut paychobogicel phate 34-1, SELINKER
success in a second language affects, as we know from observation,
4 small percentage of learners—perhaps @ mere 5% It follows from
this assumption that this 5% go through very different psycho-
linguistic processes than do most second language learners and that
these successful learners may be safely ignored—in a counterfactual
sense®—for the purposes of establishing the construets which point
to the psychologically relevant data pertinent to most second lan-
‘guage learners. Regarding the study of the latter group of learners
. the vast majority of second language learners who fail to achieve
native speaker competence), the notion of ‘attempted learning’ is
independent and logically prior to the notion of ‘successful learning’
In this paper we will focus on attempted learning by this group of
learners, successful or not, and will assume that they activate a
different, though still genetically determined structure (referred to
haere as the Jatent psychological structure) whenever they attempt to
produce a sentence in the second-language, that is, whenever they
attempt to express meanings which they may already have, in a
language which they are in the process of learning.
This series of assumptions must be made, T think, because the
second language learner who actually achieves native speaker
competence, cannot possibly have been taught this competence,
since linguists are daily—in almost every generative study—dis-
covering new and fundamental facts about particular languages.
Successful learners, in order to achieve this native-speaker com-
ppetence, must have acquired these facts (and most probably
important principles of language organization) without having
explicitly been taught thera.”
Regarding the ideal second language learner who will not ‘succeed
(in the absolute sense described above) and who is thus representative
of the vast majority of second language learners, we can idealize
that from the beginning of his study of a second language, he has
his attention focused upon one norm of the language whose sen-
tences he is attempting to produce. With this statement, we have
idealized the picture we wish to sketch in the following ways:* the
generally accepted notion ‘target language’ (TL), ie. the second
Janguage the fearner is attempting to learn, is here restricted to mean
that there is only one norm of one dialect within the interfimgwal focus
‘of attention of the learner. Furthermore, we focus our analytical
attention upon the only observable data to which we can relate theo-
retical predictions:* the utterances which are produced when the
learner attempts to say sentences of a TL. This set of utterances for
INTERLANGUAGE 35
‘most learners ofa second language is not identical to the hypothesized
corresponding set of utterances which would have been produced
bya native speaker of the TL had he attempted to express the same
‘meaning as the learner. Since we can observe that these two sets of
utterances are not identical, then in the making of constructs relevant
toa theory of second language learning. one would be completely
justified in hypothesizing, perhaps even compelled to hypothesize,
the existence ofa separate linguistic system'® based on the observable
output which results from a learner's attempted production of
TL norm. This linguistic system we wilt call “interlanguage’ (IL)
‘One of the main points of this paper is the assumption that predi
tions of behavioral events in a theory of second language learning
should be primarily concerned with the linguistic shapes of the
utterances produced in ILs. Successful predictions of such behavioral
events in meaningful performance situations will add credence to
the theoretical constructs related to the latent psychological structure
discussed in this paper.
It follows from the above that the only observable data from
meaningful performance situations we can establish as relevant 10
interlingual identifications are: (1) utterances in the learner's native
language (NL) produced by the learner; (2) IL utterances produced
by the leamer; and (3) TL utterances produced by native speakers
of that TL. These three sets of utterances or behavioral events are,
then, in this framework, the psychologically relevant data of
second language learning, and theoretical predictions in a relevant
psychology of second language learning willbe the surface structure
of IL sentences.
By setting up these three sets of utterances within one theoretical
framework, and by gathering as data utterances related to specific
Jinguistic structures in each of these three systems (under the same
‘experimental conditions, if possible) the investigator in the psychol-
ogy of second language learning can begin to study the psycho-
linguistic processes which establish the knowledge which underlies
IL behavior. 1 would like to suggest that there are five central
processes (and perhaps some additional minor ones), and that they
exist in the latent psychological structure referred to above. I con-
sider the following to be processes central to second language
earning; firs, language transfer; second, transfer of training: third,
Strategies of second language learning; fourth, strategies of second
language communication; and fifth, overgeneralization of TL
linguistic material, Each of the analyst's predictions as to the shape 36 L. SELINKER
of IL utterances should be associated with one or more of these, or
other, processes.
3 Fossilization
Before briefly describing these psycholinguistic processes, another
notion I wish to introduce for: the reader's consideration is the
concept of fossilization, a mechanism which is assumed also to exist
in the latent psychological structure described above. Fosslizable
linguistic phenomena are linguistic items, rules, and subsystems
which speakers of a particular NL will tend to keep in their IL
relative to a particular TL, no matter what the age of the learner or
amount of explanation and instruction he receives in the TL.‘ I
have in mind such fossilizable structures as the well-known ‘errors’;
French uvular /r/ in their English IL, American English retroflex /r/
in their French IL, English rhythm in the IL relative to Spanish,
German Time-Place order after the verb in the English 1L of German
speakers, and so on. I also have in mind less well known ‘non-errors’
such as Spanish monophthong vowels in the IL of Spanish speakers
relative to Hebrew, and Hebrew Object-Time surface order after the
verb in the IL of Hebrew speakers relative to English. Finally, there
are fossilizable structures that are much harder to classify such as
some features of the Thai tone system in the IL of Thai speakers
relative to English. It is important to note that fossilizable structures
tend to remain as potential performance, re-emerging'* in the
productive performance of an IL even when seemingly eradicated.
Many of these phenomena reappear in IL performance when the
learner's attention is focused upon new and difficult intellectual
subject matter or when he is ina state of anxiety or other excitement,
and strangely enough, sometimes when he is in a state of extreme
relaxation. Note that the claim is made here that, whatever the cause,
the well-observed phenomenon of “backsliding’ by second language
learners from a TL norm is not, as has been genecally believed, either
random or towards the speaker's NL, but toward an IL norm.'*
‘A crucial fact, perhaps the most crucial fact, which any adequate
theory of second language learning will have to explain is this
regular reappearance or re-emergence in IL productive perform-
ance of linguistic structures which were thought to be eradicated.
This behavioral reappearance is what has led me to postulate the
reality of fossilization and ILs. It should be made clear that the
reappearance of such behavior is not limited to the phonetic level.
INTERLANGUAGE 37
For example, some of the subtlest input information that a learner
ofa second language has to master regards subcategorization notions
‘of verbal complementation. Indian English as an IL with regard to
Englsh'* seems to fossilize the ‘that complement’ or V that con-
struction forall verbs that take sentential complements. Even when
the correct form has been learned by the Indian speaker of English,
this type of knowledge isthe first he seems to lose when his attention
is diverted to new intellectual subject matter or when he has not
spoken the TL for even a short time. Under conditions such as these,
there is a regular reappearance of the ‘that complement’ in IL per-
formance for all sentential complements.
4 Five Central Processes
It is my contention that the most interesting phenomena in IL
performance are those items, rules, and subsystems which are
fossilizable in terms of the five processes listed above. If it cdf
bbe experimentally demonstrated that fossilizable items, rules and
subsystems which occur in IL performance are a result of the NL,
then we are dealing with the process of language transfer; if these
fossilizable items, rules, and subsystems are a result of identifiable |
items in training procedures, then we are dealing with the process |
known as transfer of training:
approach by the learner to the material to be learned, then we are
dealing with strategies of second language learning; if they are a
result of an identifiable approach by the learner to communica
‘with native speakers of the TL, then we are dealing with strategies
of second language communication; anv finally, if they are a result
of a clear overgeneralization of TL rules and semantic features, then
we are dealing with the overgeneralization of TL linguistic material.
I would like to hypothesize that these five processes are processes
which are central to second language learning, and that each process
forces fossilizable material upon surface IL utterances, controlling
toa very large extent the surface structures of these utterances.
Combinations of these processes produce what we might term
entirely fossilized IL competence. Coulter (1968) presents convincing
data to demonstrate not only language transfer but also a strategy of
Commianicaffon ConmonTS thany second language learners. This.
strategy of communication dictates to them, internally as it were,
‘that they know enough of the TL in order to communicate. And they
slap learning."* Whether they stop learning entirely or go oa to learn
they are a result of an identifiable |
i 38. SELINKER
in a minor way, ¢.g. adding vocabulary as experience demands (cf
Jain) is, it seems to me, a moot point. If these individuals do not also
learn the syntactic information that goes with lexical items, then
adding a few nev. lexical items, say on space travel, is, | would argue,
of litte consequence. The important thing to note with regard to the
cevidenee presented in Coulter (1968) and Jain (1969) is that not only
ccan entire IL competences be fossilized in individual Tearners per
forming in their own interlingual situation," but also in whole
‘groups of individuals, resulting in the emergence of a new dialect
(here Indian English), where fossilized IL compete be the
normal situation. a ——
We will now provide examples of these processes. The examples
presented in section 3 are almost certainly the result of the process of|
language transfer. A few examples relating to the other processes
should suffice for this paper.
4.1 Overgeneralization of TL rules is a phenomenon well known to
Janguage teachers. Speakers of many languages coulé produce a
sentence of the following kind in their English IL,
(1) What did he intended to say?"*
Where the past tense morpheme -ed is extended to an environment
in which, to the learner, it could logically apply, but just does not.
‘The Indian speaker of English who produces the collocation drive a
bicycle in his IL performance, as in (2)
(@) After thinking litte I decided to start on the bicycle as
slowly as I could as it was not possible to drive fast.
is most probably ovecgeneralizing the use of drive to all vehicles
(ain, 1969, ef. note 26 here). Most learners of English quickly
learn the English rule of contraction which forms things like the
concert’s from the concert is, but then these learners may over-
‘generalize this rule to produce sentences like,
(3) Max is happier than Sam’s these days
in their English IL, Though this sentence is hypothetical, it illustrates
an earlier point. The learner of English who produces contractions
correctly in all environments must have learned the following
‘constraint without explanation and instruction’, since this constraint
was discovered only recently; “contraction of auxiliaries . .. cannot
‘scout when a constituent immediately following the auxiliary to be
INTERLANGUAGE 39
contracted has been deleted,” e.g. ‘happy" in (3) (Lakoff, in press).
Dozens of examples of overgeneralization of TL rules are provided
in Richards (1971).
4.2 The transfer of training is a process which is quite different from
language transfer (see Selinker, 1969) and from overgeneralization
of TL rules. It underlies the Source of a difficulty which Serbo-
Croation speakers at all levels of English proficiency regularly have
with the he/she distinction, producing in their English IL he on
almost every occasion wherever he or she would be called for accord-
ing to any norm of English. There is no language transfer effect
here since, with regard to animateness, the distinction between he
and she is the same in Serbo-Croatian as it is in English.'® According
to a standard contrastive analysis then there should be no trouble.
Itseems to be the case that the resultant IL form, in the first instance,
is due directly to the transfer of training; textbooks and teachers in
this interlingual situation almost always present drills with he and
never with she. The extent of this fossilization can be seen with
respect to the speakers ofthis IL over the age of 18, who even though
they are consciously aware of the distinction and of their recurrent
error, in fact, regularly produce he for both he and she, stating that
they feel they do not need to make this distinction in order to com-
municate.2® In this case, then, the fossilizable error is due originally
toa type of transfer of training and later to a particular strategy of
second language communication.
4.3 Concerning the notion ‘strategy’ little is known in psychology
about what constitutes a strategy, and a viable definition of it does
rot seem possible at present. Even less is known about strategies
which leasners of a second language use in their attempt to master a
‘TLand express meanings init. Ithas been pointed out" that learner
strategies are probably culturé-botind to some extent. For example,
in many traditional cultures, chanting is used as a learning device,
clearly relating to what is learned in those situations. Crucially, it
has been argued” that strategies for handling TL linguistic material
evolve whenever the learner realizes, either consciously or sub-
consciously, that he has no linguistic competence with regard to some
aspect of the TL. It cannot be doubted that various internal strate-
‘gies? on the part of the second language learner affect to a large
extent the surface structure of sentences underlying IL utterances.
But exactly what these strategies might be and how they might work 40. L. SELINKER
is at present pure conjecture. Thus, one can only roughly attribute
the source of the examples presented herein 10 one or another
strategy.
‘One example of a strategy of second language learning that is
‘widespread in many interlingual situations is a tendency on the part
“of earness o reduce the TL to a simpler system. According to Jain
(1969) the results of this strategy are manifested at all levels of syntax
in the IL of Indian speakers of English. For example, if the learner
has adopted the strategy that all verbs ave either transitive or intran-
sitive, he may produce IL forms such as
(4) Lam feeling thirsty
or (5) Don’t worry, I’m hearing him
and in producing them they seem to have adopted the further
strategy that the realization of the category ‘aspect’ in its progressive
form on the surface is always with ing marking for further discussion,
see Jain, 1969).
Coulter (1968) reports systematic errors occurring in the English
IL performance of two elderly Russian speakers of English, due to
another strategy which seems also to be widespread in many
interlingual situations; a tendency on the part of second language
learners to avoid grammatical formatives such as articles (6),
plural forms (7), and past tense forms (8);
(6) It was @ nice, nice trailer, 9 big one. (Coulter, 1968, 22)
(7) Uhave many hundred carpenter my own (ibid, 29)
(8) I was in Frankfurt when I fill application (ibid, 36).
This tendency could be the result of a learning strategy of simpli-
fication, but Coulter (1968) attributes it to a communication strategy
due to the past experience of the speaker which has shown him that
ithe thinks about grammatical processes while attempting to express
in English meanings which he already has, then his speech will be
hesitant and disconnected, leading native speakers to be impatient
with him. Also, Coulter claims that this strategy of second language
communication seemed to dictate to these speakers that a form such
as the English plural ‘was not necessary for the kind of communicat-
ing they used’ (ibid, p. 30)
Not all of these strategies, it must be pointed out, are conscious.
A subeonscious strategy of second language learning called “cue
‘copying’ has beenexperimented with by Crothers and Suppes
CHGS on Areca ering Rustin morhologil eoncep
INTERLANGUAGE 41
‘This ‘copy the cue’ strategy is most probably due to what they call
“probabilify matching’, where the chance that the learner will select
‘an alternative mozphological ending related to the cue noun is not
random. Crothers and Suppes do not provide examples of the result
of this strategy in meaningful performance situations; an example
would be the r at the end of words like California and. saw wich
foreign students of English who have had teachers from the Boston
area regularly reproduce in their English IL.
44 To conclude this section, it should be pointed out that beyond
the five so-called central processes, there exist many other processes
which account to some degree for the surface forms of IL utterances.
ne might mention spelling pronunciations, e.g. speakers of many
‘anguages pronounce final -er on English words as [é] plus some form
ofr; cognate pronunciation, e.g. English athlete pronounced as
[atlit} by many Frenchmen whether or not they can produce (8) in
other English words;"* holophrase learning (Jain, 1969), e.g. from
half-an-hour the Indian learner of English may produce one and
half-an-hour; hypercorrection, e.@. the Israeli who in attempting to
‘get rid of his uvular fricative for English retroflex {r} produces [w]
before front vowels, ‘a vocalization too far forward’?*; and most
assuredly others such as long exposure fo signs and headlines which
ccording to Jain (1969) affect by themselves the shape of English
IL utterances of Indians, or at least reinforce some important pro-
cesses such as language transfer.
5 Problems with this perspective
There are certainly many questions one might wish to ask regarding
the perspective presented so far im this paper; I shall attempt to deal
with five (5.1-5.5). The reader should bear in mind that we are here
calling for the discovery, description and experimental testing of
fossilizable items, rules and subsystems in interlanguages ana the
relating of these to the above mentioned processes—especially to the
central ones, What seems to be most promising for study is the
observation concerning fossilization, Many TL linguistic structures
are never really eradicated for most second language learners;
manifestations of these structures regularly reappear in IL. pro-